CN112560719B - High-resolution image water body extraction method based on multi-scale convolution-multi-core pooling - Google Patents
High-resolution image water body extraction method based on multi-scale convolution-multi-core pooling Download PDFInfo
- Publication number
- CN112560719B CN112560719B CN202011518292.XA CN202011518292A CN112560719B CN 112560719 B CN112560719 B CN 112560719B CN 202011518292 A CN202011518292 A CN 202011518292A CN 112560719 B CN112560719 B CN 112560719B
- Authority
- CN
- China
- Prior art keywords
- water body
- pooling
- image
- core
- scale convolution
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/28—Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/30—Assessment of water resources
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Bioinformatics & Computational Biology (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Evolutionary Biology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Astronomy & Astrophysics (AREA)
- Remote Sensing (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a multi-scale convolution-multi-core pooling-based high-resolution image water body extraction method, which comprises the steps of obtaining a target image, preprocessing the target image based on a self-adaptive histogram equalization algorithm, obtaining an updated target image, inputting the updated target image into a trained multi-scale convolution-multi-core pooling coding-decoding network model, and extracting a water body region in the target image; the water body region extraction model is a multi-scale convolution-multi-core pooling coding-decoding network model. The method provided by the invention can reduce the feature loss of the original image, and can be used for extracting various scale water bodies from the high-resolution image with high precision, high accuracy and automation.
Description
Technical Field
The invention relates to the technical field of water body extraction, in particular to a high-resolution image water body extraction method based on multi-scale convolution-multi-core pooling.
Background
The water resource plays an important role in human society, the protection utilization of the water resource is enhanced, and the detection and management are important. Therefore, the space-time distribution of the water body is researched, and the accurate and efficient extraction of the water body has positive effects on resource investigation, urban planning and flood control and disaster reduction. The rapid development of computer technology and the availability of a large amount of remote sensing data lay a foundation for extracting water body by using remote sensing images.
The traditional method for extracting the water body can be classified into a 3-class exponential method, an image transformation method and a machine learning method. The method has the following problems that (1) a threshold value needs to be set manually when water is extracted, the dependence on knowledge and experience of people is strong (2) the spatial information of a remote sensing image is ignored, and the information is lost (3) the fine water with shadow and sediment is difficult to extract.
Currently, a convolutional neural network method in deep learning is widely applied to water extraction, and Kang et al propose a CNNs model for extracting water based on high-resolution satellite images and combining U-net and Densenet structures in a remote sensing image water extraction method and system based on deep learning (CN 109934095A, 2019-06-25); zeng Anming et al propose a method for extracting a water body from an input remote sensing image by using a pretrained convolutional neural network in a high-resolution remote sensing image water body extraction method based on multiscale optimization (CN 110427836A, 2019-11-08), wherein the multiscale feature map obtains a first-trial rough water body segmentation result through a first classifier, and water body extraction is realized through iterative optimization.
Although the two methods realize the extraction of the remote sensing image water body, the following problems exist: (1) in a convolutional neural network model, in the forward propagation process, continuous convolution and pooling operation can lose image characteristics, so that the boundary of a part of regional water body is blurred; (2) adjacent ground objects in the image have similar spectral characteristics, large-area water areas and tiny water bodies are unevenly distributed, convolution kernels and pooling kernels with fixed sizes easily cause water body information loss, and characteristics cannot be extracted from multiple angles, so that the problem of low accuracy of water body extraction results is generated.
Disclosure of Invention
The purpose of the invention is that: the method for extracting the water body from the image is capable of reducing the loss of the original image characteristics and is high in accuracy of extracting the water body.
The technical scheme is as follows: the invention provides a multi-scale convolution-multi-core pooling-based high-resolution image water body extraction method, which is used for extracting a water body region in a target image and is characterized by comprising the following steps of:
step 2, preprocessing the target image based on a self-adaptive histogram equalization algorithm to obtain an updated target image;
step 3, inputting the updated target image into a trained water body region extraction model, and extracting a water body region in the target image; the water body region extraction model is a multi-scale convolution-multi-core pooling coding-decoding network model.
As a preferred scheme of the invention, the multi-scale convolution-multi-core pooling coding-decoding network model comprises an encoder, a decoder and a bridging module;
the encoder adopts a Residual block and a pooling layer to perform downsampling; the decoder performs up-sampling by using a transposed convolutional layer; the bridging module comprises a multi-scale convolution block and a multi-core pooling block;
the bridging module bridges the encoder and decoder bottoms through a multi-scale convolution block and a multi-core pooling block therein.
As a preferred embodiment of the present invention, the Residual block adopts the BasicbLock structure of ResNet 18/34.
As a preferred scheme of the present invention, the multi-scale convolution block includes at least one hole convolution string, expansion rates of the hole convolution strings are different from each other, and the hole convolutions are connected in series-parallel; the multi-scale convolution block is according to the following formula:
X 1 ＝r×(k-1)+1 (1)
X m ＝r×(k-1)+X m-1 (2)
obtaining receptive fields, wherein X 1 Is the size of the first receptive field, X m Represents the size of the receptive field of the m layer, X m-1 The size of the receptive field of the upper layer; k is the size of the cavity convolution kernel; and r is the expansion rate.
As a preferable scheme of the invention, the multi-core pooling block comprises at least one pooling block, the pooling cores of the pooling blocks are different in size, and the pooling blocks are connected in parallel; and the multi-core pooling block performs up-sampling by adopting a bilinear interpolation method.
As a preferred aspect of the present invention, the method further comprises training a multi-scale convolutional-multi-core pooled encoding-decoding network model by the method described below:
step A, acquiring an image data set, wherein the image data set comprises a plurality of images and labels corresponding to each pixel in each image; the labels are segmentation categories of pixels in the image, and the segmentation categories comprise water body categories;
step B, preprocessing each image in the image data set based on a self-adaptive histogram equalization algorithm;
step C, performing binarization segmentation on the labels in the original image data set, and dividing the labels into a water body and a non-water body, so as to obtain a water body prospect classification map; the original image data set is an image data set which is not preprocessed;
step D, carrying out data expansion on the preprocessed images and the water body foreground two-class diagram, and further constructing a water body extraction training data set; and E, building a multi-scale convolution-multi-core pooling coding-decoding network model, and training the model by using a water body extraction training data set to obtain a trained multi-scale convolution-multi-core pooling coding-decoding network model.
As a preferred embodiment of the present invention, in step D, the data expansion method includes: and carrying out non-overlapping clipping, turnover transformation, random rotation, gamma transformation and noise addition on each image and the corresponding water body foreground two-class diagram.
The beneficial effects are that: compared with the prior art, the method provided by the invention can be used for extracting various scale water bodies from high-resolution remote sensing images in a high-precision and automatic manner through the coding-decoding convolutional neural network model based on multi-scale convolution-multi-core pooling.
Drawings
FIG. 1 is a flow chart of a water extraction method provided according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of an image preprocessing result according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of another image preprocessing result according to an embodiment of the present invention;
FIG. 4 is a diagram of a water foreground classification tag according to the mask extraction of an original image and an original tag provided by an embodiment of the present invention;
fig. 5 is a diagram illustrating a structure of an encoding-decoding network model according to an embodiment of the present invention.
Detailed Description
The invention is further described below with reference to the accompanying drawings. The following examples are only for more clearly illustrating the technical aspects of the present invention, and are not intended to limit the scope of the present invention.
Referring to fig. 1, the method provided by the invention comprises the following specific steps:
step 2, preprocessing the target image based on a self-adaptive histogram equalization algorithm to obtain an updated target image;
step 3, inputting the updated target image into a trained water body region extraction model, and extracting a water body region in the target image; the water body region extraction model is a multi-scale convolution-multi-core pooling coding-decoding network model.
The multi-scale convolution-multi-core pooling coding-decoding network model comprises a coder, a decoder and a bridging module; the encoder adopts a Residual block and a pooling layer to perform downsampling; the decoder performs up-sampling by using a transposed convolutional layer; the bridging module comprises a multi-scale convolution block and a multi-core pooling block; the Residual block adopts the BasicbLock structure of ResNet 18/34.
The bridging module bridges the encoder and decoder bottoms through the multi-scale convolution blocks and the multi-core pooling blocks therein.
The multi-scale convolution block comprises at least one cavity convolution string, the expansion rates of the cavity convolution strings are different from each other, and the cavity convolution strings are connected in parallel; the multi-scale convolution block is according to the following formula:
X 1 ＝r×(k-1)+1 (1)
X m ＝r×(k-1)+X m-1 (2)
obtaining receptive fields, wherein X 1 Is the size of the first receptive field, X m Representing the size of receptive field, X m-1 The size of the receptive field of the upper layer; k is the size of the cavity convolution kernel; r is the expansion rate; .
The multi-core pooling block comprises at least one pooling block, the pooling cores of the pooling blocks are different in size, and the pooling blocks are connected in parallel; and the multi-core pooling block performs up-sampling by adopting a bilinear interpolation method. The multi-core pooling block adopts a structure similar to pyramid pooling, pools with different sizes are connected in parallel, and after convolution dimension reduction treatment of 1 multiplied by 1, feature fusion is carried out, and interpolation is carried out in a bilinear interpolation mode. The method for training the multi-scale convolution-multi-core pooling coding-decoding network model comprises the following steps:
step A, acquiring an image data set, wherein the image data set is a high-resolution remote sensing image data set, and the image data set comprises a plurality of images and labels corresponding to pixels in the images; the labels are segmentation categories of pixels in the image, and the segmentation categories comprise water body categories;
the high-resolution image data set refers to an open-source high-resolution image semantic segmentation data set obtained from the Internet, and a water body type exists in the image segmentation type of the data set; there are a number of open-sourced high-resolution remote sensing image datasets in the internet, where there are water categories in the segmentation category, such as the land cover classification dataset deep globe (i.demir et al deep 2018).
And B, preprocessing the remote sensing images in the high-resolution remote sensing image dataset, and enhancing the image contrast.
The pretreatment method comprises the following steps: each image in the image dataset is preprocessed based on an adaptive histogram equalization algorithm.
Schematic diagrams of the pretreatment results are shown in fig. 2 and 3.
Step C, performing binarization segmentation on the labels in the original image data set, and dividing the labels into a water body and a non-water body, so as to obtain a water body prospect classification map; the original image dataset is an image dataset that has not been preprocessed.
In one embodiment, the acquired water foreground two-class diagram is shown in fig. 4.
And D, performing data expansion on the preprocessed images and the water foreground two-class diagram, further constructing a water extraction data set, and dividing the data in the data set into a training data set and a testing data set according to a preset proportion.
And B, processing image data in the original data set, C, processing labels to obtain required labels, and D, ensuring one-to-one correspondence between the images and the labels.
The data expansion method comprises the following steps: and carrying out non-overlapping clipping, turnover transformation, random rotation, gamma transformation and noise addition on each image and the corresponding water body foreground two-class diagram.
By using the data set expansion method, a water body classification data set is manufactured, and images and labels in the data set are subjected to non-overlapping cutting, overturning transformation, random rotation, gamma transformation and noise addition, so that a large number of images with fixed sizes are generated, the robustness of the model is improved, and the over fitting of training data is prevented.
E, building a multi-scale convolution-multi-core pooling coding-decoding network model, training the model by using the water body extraction training data set obtained in the step A to the step D, and storing weight parameters; and importing the weight parameters and the test data set into a multi-scale convolution-multi-core pooling coding-decoding network model, testing the model, and reserving the weight parameters with the best test results.
A multi-scale convolution-multi-core pooling network model with the encoding-decoding structure as shown in fig. 4 is constructed. Wherein the encoder downsamples with residual blocks of the basic cblock structure of ResNet 18/34; the decoder upsamples with a transposed convolutional layer; the multi-scale hole convolution block and the multi-core pooling block with improved pyramid pooling structure bridge between the encoder and the decoder. And loading the training data set into the built neural network for training, and reserving the weight parameter with the best water body extraction result after the test of the test set.
The method provided by the invention adopts the coding-decoding network structure of the residual block for downsampling, effectively solves the problem of image characteristic loss caused by continuous rolling and pooling operation, and improves the extraction precision of the boundary of partial regional water body; the multi-scale cavity convolution is used, the receptive field is increased, the precision of extracting the fine water body is more effectively improved, the pyramid-pooling multi-core pooling structure is improved, the characteristics are extracted and aggregated from different angles, and the reduction of the water body extraction precision caused by the spectrum similarity of adjacent ground objects is avoided; by utilizing the existing high-resolution image data set, the time and labor consuming manual labeling of training labels is avoided, and meanwhile, comparison of different methods under the same data set is facilitated.
The foregoing is merely a preferred embodiment of the present invention, and it will be apparent to those skilled in the art that modifications and variations can be made without departing from the technical principles of the present invention, and the modifications and variations should also be regarded as the scope of the invention.
Claims (3)
1. The high-resolution image water body extraction method based on multi-scale convolution-multi-core pooling is used for extracting a water body region in a target image and is characterized by comprising the following steps of:
step 1, obtaining a target image;
step 2, preprocessing the target image based on a self-adaptive histogram equalization algorithm to obtain an updated target image;
step 3, inputting the updated target image into a trained water body region extraction model, and extracting a water body region in the target image; the water body region extraction model is a multi-scale convolution-multi-core pooling coding-decoding network model;
the multi-scale convolution-multi-core pooling coding-decoding network model comprises a coder, a decoder and a bridging module;
the encoder adopts a Residual block and a pooling layer to perform downsampling; the decoder performs up-sampling by using a transposed convolutional layer; the bridging module comprises a multi-scale convolution block and a multi-core pooling block; the Residual block adopts a BasicbLock structure of ResNet 18/34;
the bridging module bridges the bottom ends of the encoder and the decoder through a multi-scale convolution block and a multi-core pooling block;
the multi-scale convolution block comprises at least one cavity convolution string, the expansion rates of the cavity convolution strings are different from each other, and the cavity convolution strings are connected in parallel; the multi-scale convolution block is according to the following formula:
X 1 ＝r×(k-1)+1 (1)
X m ＝r×(k-1)+X m-1 (2)
obtaining receptive fields, wherein X 1 Is the size of the first receptive field, X m Represents the size of the receptive field of the m layer, X m-1 The size of the receptive field of the upper layer; k is the size of the cavity convolution kernel; r is the expansion rate;
the multi-core pooling block comprises at least one pooling block, the pooling cores of the pooling blocks are different in size, and the pooling blocks are connected in parallel; and the multi-core pooling block performs up-sampling by adopting a bilinear interpolation method.
2. The multi-scale convolution-multi-core pooling-based high-resolution image water extraction method according to claim 1, further comprising training a multi-scale convolution-multi-core pooling encoding-decoding network model by:
step A, acquiring an original image data set, wherein the data set comprises a plurality of images and labels corresponding to each pixel in each image; the labels are segmentation categories of pixels in the image, and the segmentation categories comprise water body categories;
step B, preprocessing each image in the original image dataset based on a self-adaptive histogram equalization algorithm;
step C, performing binarization segmentation on the labels in the original image data set, and dividing the labels into a water body and a non-water body, so as to obtain a water body prospect classification map; the original image data set is an image data set which is not preprocessed;
step D, carrying out data expansion on the preprocessed images and the water body foreground two-class diagram, and further constructing a water body extraction data set;
and E, building a multi-scale convolution-multi-core pooling coding-decoding network model, and training the model by using a water body extraction training data set to obtain a trained multi-scale convolution-multi-core pooling coding-decoding network model.
3. The multi-scale convolution-multi-core pooling-based high-resolution image water extraction method according to claim 2, wherein in step D, the data expansion method comprises: and carrying out non-overlapping clipping, turnover transformation, random rotation, gamma transformation and noise addition on each image and the corresponding water body foreground two-class diagram.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011518292.XA CN112560719B (en) | 2020-12-21 | 2020-12-21 | High-resolution image water body extraction method based on multi-scale convolution-multi-core pooling |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011518292.XA CN112560719B (en) | 2020-12-21 | 2020-12-21 | High-resolution image water body extraction method based on multi-scale convolution-multi-core pooling |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112560719A CN112560719A (en) | 2021-03-26 |
| CN112560719B true CN112560719B (en) | 2023-07-04 |
Family
ID=75031135
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011518292.XA Active CN112560719B (en) | 2020-12-21 | 2020-12-21 | High-resolution image water body extraction method based on multi-scale convolution-multi-core pooling |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112560719B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113343861B (en) * | 2021-06-11 | 2023-09-05 | 浙江大学 | Remote sensing image water body region extraction method based on neural network model |
| CN114461837A (en) * | 2022-02-16 | 2022-05-10 | 北京百度网讯科技有限公司 | Image processing method and device and electronic equipment |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6269192B1 (en) * | 1997-07-11 | 2001-07-31 | Sarnoff Corporation | Apparatus and method for multiscale zerotree entropy encoding |
| CN110287962A (en) * | 2019-05-20 | 2019-09-27 | 平安科技（深圳）有限公司 | Remote Sensing Target extracting method, device and medium based on superobject information |
| CN110781775A (en) * | 2019-10-10 | 2020-02-11 | 武汉大学 | Remote sensing image water body information accurate segmentation method supported by multi-scale features |
| CN111079649A (en) * | 2019-12-17 | 2020-04-28 | 西安电子科技大学 | Remote sensing image ground feature classification method based on lightweight semantic segmentation network |
| CN111127493A (en) * | 2019-11-12 | 2020-05-08 | 中国矿业大学 | Remote sensing image semantic segmentation method based on attention multi-scale feature fusion |
| CN111563491A (en) * | 2020-07-14 | 2020-08-21 | 浙江大华技术股份有限公司 | Method, equipment and device for segmenting remote sensing image by using network model |
| CN111783782A (en) * | 2020-05-29 | 2020-10-16 | 河海大学 | Remote sensing image semantic segmentation method fusing and improving UNet and SegNet |
| CN112001293A (en) * | 2020-08-19 | 2020-11-27 | 四创科技有限公司 | Remote sensing image ground object classification method combining multi-scale information and coding and decoding network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11391819B2 (en) * | 2018-07-18 | 2022-07-19 | Qualcomm Incorporate | Object verification using radar images |
-
2020
- 2020-12-21 CN CN202011518292.XA patent/CN112560719B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6269192B1 (en) * | 1997-07-11 | 2001-07-31 | Sarnoff Corporation | Apparatus and method for multiscale zerotree entropy encoding |
| CN110287962A (en) * | 2019-05-20 | 2019-09-27 | 平安科技（深圳）有限公司 | Remote Sensing Target extracting method, device and medium based on superobject information |
| CN110781775A (en) * | 2019-10-10 | 2020-02-11 | 武汉大学 | Remote sensing image water body information accurate segmentation method supported by multi-scale features |
| CN111127493A (en) * | 2019-11-12 | 2020-05-08 | 中国矿业大学 | Remote sensing image semantic segmentation method based on attention multi-scale feature fusion |
| CN111079649A (en) * | 2019-12-17 | 2020-04-28 | 西安电子科技大学 | Remote sensing image ground feature classification method based on lightweight semantic segmentation network |
| CN111783782A (en) * | 2020-05-29 | 2020-10-16 | 河海大学 | Remote sensing image semantic segmentation method fusing and improving UNet and SegNet |
| CN111563491A (en) * | 2020-07-14 | 2020-08-21 | 浙江大华技术股份有限公司 | Method, equipment and device for segmenting remote sensing image by using network model |
| CN112001293A (en) * | 2020-08-19 | 2020-11-27 | 四创科技有限公司 | Remote sensing image ground object classification method combining multi-scale information and coding and decoding network |
Non-Patent Citations (5)
| Title |
|---|
| Automatic Water-Body Segmentation From High-Resolution Satellite Images via Deep Networks;Ziming Miao 等;《IEEE GEOSCIENCE AND REMOTE SENSING LETTERS 》;第15卷(第4期);1-5 * |
| Multi-scale context extractor network for water-body extraction from high-resolution optical remotely sensed images;Kang Jian 等;《International Journal of Applied Earth Observation and Geoinformation》;第103卷;1-12 * |
| 基于卷积神经网络的高分二号遥感影像提取方法;魏青迪;《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》(第(2020)01期);C028-163 * |
| 基于多尺度特征增强网络的高分辨遥感影像水体提取;康健;《中国优秀硕士学位论文全文数据库基础科学辑》(第(2023)01期);A008-304 * |
| 基于深度学习的高分辨率遥感图像建筑物识别;宋廷强 等;《计算机工程与应用》;第56卷(第8期);26-34 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112560719A (en) | 2021-03-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110322495B (en) | Scene text segmentation method based on weak supervised deep learning | |
| CN110136154B (en) | Remote sensing image semantic segmentation method based on full convolution network and morphological processing | |
| CN108921799B (en) | Remote sensing image thin cloud removing method based on multi-scale collaborative learning convolutional neural network | |
| CN110889449A (en) | Edge-enhanced multi-scale remote sensing image building semantic feature extraction method | |
| CN114120102A (en) | Boundary-optimized remote sensing image semantic segmentation method, device, equipment and medium | |
| CN112668494A (en) | Small sample change detection method based on multi-scale feature extraction | |
| CN112183258A (en) | Remote sensing image road segmentation method based on context information and attention mechanism | |
| CN112132149B (en) | Semantic segmentation method and device for remote sensing image | |
| CN111079683A (en) | Remote sensing image cloud and snow detection method based on convolutional neural network | |
| CN113780149A (en) | Method for efficiently extracting building target of remote sensing image based on attention mechanism | |
| CN112560719B (en) | High-resolution image water body extraction method based on multi-scale convolution-multi-core pooling | |
| CN114187450A (en) | Remote sensing image semantic segmentation method based on deep learning | |
| CN115661144A (en) | Self-adaptive medical image segmentation method based on deformable U-Net | |
| CN111583276B (en) | CGAN-based space target ISAR image component segmentation method | |
| CN112101364B (en) | Semantic segmentation method based on parameter importance increment learning | |
| CN112836614B (en) | High-resolution remote sensing image classification method based on residual error network and transfer learning | |
| CN114067019A (en) | Urban waterlogging risk map rapid prefabricating method coupling deep learning and numerical simulation | |
| CN112001293A (en) | Remote sensing image ground object classification method combining multi-scale information and coding and decoding network | |
| CN112700418A (en) | Crack detection method based on improved coding and decoding network model | |
| CN116958827A (en) | Deep learning-based abandoned land area extraction method | |
| CN112837320A (en) | Remote sensing image semantic segmentation method based on parallel hole convolution | |
| CN110472632B (en) | Character segmentation method and device based on character features and computer storage medium | |
| CN115330703A (en) | Remote sensing image cloud and cloud shadow detection method based on context information fusion | |
| CN116129280B (en) | Method for detecting snow in remote sensing image | |
| CN114612315A (en) | High-resolution image missing region reconstruction method based on multi-task learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |